Download (ALLHAT) heart failure Validation Study

Prevention and Rehabilitation
The Antihypertensive and Lipid Lowering Treatment to
Prevent Heart Attack Trial (ALLHAT) heart failure
Validation Study: Diagnosis and prognosis
Paula T. Einhorn, MD, MS,a Barry R. Davis, MD, PhD,b Barry M. Massie, MD,c William C. Cushman, MD,d
Linda B. Piller, MD, MPH,b Lara M. Simpson, PhD,b Daniel Levy, MD,e Chuke E. Nwachuku, DrPH, MA, MPH,a
and Henry R. Black, MD,f for the ALLHAT Collaborative Research Group Bethesda, MD; Houston, TX; San
Francisco, CA; Memphis, TN; Framingham, MA; and Chicago, IL
Background ALLHAT, a randomized, double-blind, active-controlled hypertension treatment trial in 42 418 patients,
reported that a thiazide-type diuretic (chlorthalidone) was superior to a calcium channel blocker (amlodipine), an
angiotensin-converting enzyme inhibitor (lisinopril), and an a1-blocker (doxazosin) in preventing the new onset of heart
failure (HF). However, questions have been raised regarding the validity of the HF diagnosis.
Methods The ALLHAT HF Validation Study was designed to validate and elucidate the significance of HF events in
ALLHAT. Records for 2778 HF hospitalizations in 1935 patients were centrally reviewed using several prespecified
algorithms (based on ALLHAT and Framingham criteria) and reviewers’ global clinical judgment. Percent agreement with
diagnoses assigned by ALLHAT site physicians, relative risks across randomized comparisons, incidence rates, and mortality
after HF hospitalization were evaluated for first events validated by each of the criteria sets.
Results Percent agreements with site physician diagnoses were 71%, 80%, and 84% for ALLHAT, Framingham, and
reviewers’ judgment, respectively. Using these 3 criteria, relative risks (95% CI) for new-onset HF compared with
chlorthalidone were, respectively, 1.46 (1.27-1.68), 1.42 (1.25-1.62), and 1.45 (1.28-1.64) for amlodipine;
1.18 (1.02-1.28), 1.13 (0.99-1.30), and 1.15 (1.01-1.32) for lisinopril; and 1.79 (1.51-2.11), 1.71 (1.46-2.00), and
1.80 (1.55-2.10) for doxazosin.
Conclusions An independent review of source documentation showed a high degree of agreement with the HF
diagnoses assigned by site physicians and confirmed the higher risk of HF associated with first-step therapy using amlodipine,
lisinopril, or doxazosin compared with chlorthalidone. Thiazide-type diuretics should be the preferred first-step therapy for
prevention of HF in high-risk patients with hypertension. (Am Heart J 2007;153:42253.)
From the aNational Heart, Lung, and Blood Institute, Bethesda, MD, bUniversity of Texas
Health Science Center School of Public Health, Houston, TX, cSan Francisco Veterans
Affairs Medical Center, San Francisco, CA, dMemphis Veterans Affairs Medical Center,
Memphis, TN, eFramingham Heart Study/National Heart, Lung, and Blood Institute,
Framingham, MA, and fRush Presbyterian—St. Luke’s Medical Center, Chicago, IL.
This study was supported by contract NO1-HC-35130 from the National Heart, Lung,
and Blood Institute. The ALLHAT investigators acknowledge contributions of study
medications supplied by Pfizer Inc (amlodipine and doxazosin), AstraZeneca (atenolol
and lisinopril), and Bristol-Myers Squibb (pravastatin) and financial support provided by
Pfizer Inc. The National Heart, Lung, and Blood Institute sponsored the study and was
involved in all aspects other than direct operations of the study centers. This included
collection, analysis, and interpretation of the data plus the decision to submit the
manuscript for publication. Pfizer Inc, AstraZeneca, and Bristol-Myers Squibb had no role
in the design and conduct of the study; the collection, analysis, and interpretation of the
data; or the preparation or approval of the manuscript.
Disclosures. Dr Davis—consulting: GlaxoSmithKline, Merck, Takeda. Dr Massie—grant:
Bristol-Myers Squibb; consulting: Abbott, Bristol-Myers Squibb, Sanofi-Synthelabo,
GlaxoSmithKline, Novartis. Dr Cushman—grant: Abbott Laboratories, AstraZeneca,
Novartis; consulting: AstraZeneca, Bristol-Myers Squibb, Forest Pharmaceuticals, Novartis, Pfizer, Sankyo, Sanofi-Synthelabo. Dr Nwachuku—currently employed by AstraZe-
neca. Dr Black—consulting: Biovail, Boehringer Ingleheim, Bristol-Myers Squibb, CV
Therapeutics, Merck, Myogen, Novartis, Pfizer, Sanofi-Synthelabo; Data and Safety
Monitoring Board chair, Novartis (ACCOMPLISH). Dr Einhorn, Dr Piller, Dr Simpson, and
Dr Levy—none.
The ALLHAT HF Validation Study was designed by the ALLHAT Heart Failure Advisory
Group: William Cushman (Chair), John Kostis, Daniel Levy, Barry Massie, Clinical Trial
Coordinating Center (Barry Davis, Linda Piller, Sara Pressel, Lara Simpson), and The
National Heart, Lung, and Blood Institute Project Office (Paula Einhorn, Jeffrey Cutler,
Chuke Nwachuku, Michael Proschan). The ALLHAT Data and Safety Monitoring Board,
under the chairmanship of Dr Robert Califf, has reviewed the concept and procedures of
the HF validation study. A list of the ALLHAT Collaborative Research Group members has
been published previously.10
Submitted February 13, 2006; accepted October 9, 2006.
Reprint requests: Paula T. Einhorn, MD, MS, National Heart, Lung, and Blood Institute,
Division of Epidemiology and Clinical Applications, Two Rockledge Centre Room 8122,
6701 Rockledge Drive MSC 7936, Bethesda, MD 20892-7936 (20817 for courier).
E-mail: [email protected]
0002-8703/$ - see front matter
n 2007, Published by Mosby, Inc.
doi:10.1016/j.ahj.2006.10.012
American Heart Journal
Volume 153, Number 1
Einhorn et al 43
Figure 1
ALLHAT HF Validation Study population.
Heart failure (HF) is an important public health
problem. It is the most frequent cause of hospitalization
among people z65 years, and these hospitalizations are
central to the enormous cost of the disease.1,2 Despite
recently reported favorable trends in HF incidence and
survival after the onset of HF, the postincident event
mortality rates remain high (N20% at 1 year), emphasizing the importance of prevention.3,4 Although several
risk factors for HF have been identified, the Framingham
Heart Study showed that hypertension was the most
common modifiable risk factor for HF, antedating it in
91% of cases.5 Clinical trials have documented that
hypertension treatment can substantially reduce the
incidence of HF, especially in older patients with systolic
hypertension.6-8
ALLHAT, a randomized, double-blind, practice-based
hypertension treatment trial sponsored by the National
Heart, Lung, and Blood Institute, was carried out in
42 418 high-risk individuals with hypertension, but without symptomatic HF or known left ventricular ejection
fraction (LVEF) b35%.9 Heart failure was a prespecified
component of the combined cardiovascular disease (CVD)
American Heart Journal
January 2007
44 Einhorn et al
Figure 2
ALLHAT and Framingham HF criteria.
outcome, which was a major secondary end point of the
trial. Compared with chlorthalidone-based treatment, newonset HF occurred more frequently in patients randomized
to amlodipine-, lisinopril-, and doxazosin-based strategies
(relative risks [RRs] 1.38, 1.19, and 1.80, respectively).10,11
However, questions have been raised regarding the validity
of the diagnosis of HF because it was not routinely
adjudicated by the end point subcommittee and because
nonspecific peripheral edema, either occurring spontaneously or as a side effect of amlodipine, might have
prompted a misdiagnosis of HF.12-15
We undertook a validation study to systematically
evaluate all hospitalized HF events in a rigorous manner,
blinded to treatment assignment—in accordance with a
recommendation of the ALLHAT Data and Safety Monitoring Board. Based on the subset of incident HF events
with a hospitalization or followed by a HF hospitalization, this article (1) explores the agreement of HF
diagnoses assigned by ALLHAT site physicians with
6 different algorithms for HF and with the opinion of a
clinician-reviewer; (2) compares the relative risks of
validated HF between randomized treatment groups
with the relative risks previously reported in the trial’s
main results; and (3) evaluates incidence of validated HF
and examines subsequent mortality rates as indicators of
the clinical significance of this outcome.
Methods
ALLHAT design and participants
ALLHAT was a large, randomized, double-blind, activecontrolled hypertension treatment trial.9 Its overall purpose was
to determine whether newer antihypertensive drug classes—
calcium-channel blockers (amlodipine), angiotensin-converting
enzyme (ACE) inhibitors (lisinopril), or a1-blockers (doxazosin)—are superior to an older class of thiazide-type diuretics
(chlorthalidone) in preventing cardiovascular complications of
hypertension when each drug is used as the primary antihypertensive treatment (with step-up drugs as needed). The
American Heart Journal
Volume 153, Number 1
Einhorn et al 45
Table I. Baseline characteristics
Baseline characteristics
(n [%])
All HF*
(n = 2772)
Hospitalized HF
(n = 2207)
Validated HFy
y
(n = 1789)
Not validated HFz
z
(n = 146)
Age (y) (mean [SD])
55-64
z65
Black
Women
Education (y) (mean [SD])
Treated (antihypertensive)
Eligibility risk factorst
Current smoker
ASCVDt
History of MI or stroke
Type II diabetes
LVH by ECG (reported)
LVH by echo (reported)
History of CHD
BMI (mean [SD])
Blood pressure mmHg
SBP (mean [SD])
DBP (mean [SD])
Treated
SBP (mean [SD])
DBP (mean [SD])
Untreated
SBP (mean [SD])
DBP (mean [SD])
Fasting glucose (mean [SD])
GFR (mean [SD])
LDL (mg/dL) (mean [SD])
HDL (mg/dL) (mean [SD])
Triglyceride (mg/dL) (mean [SD])
70.2
750
2021
959
1198
10.9
2590
(8.1)
(27.1)
(72.9)
(34.6)
(43.2)
(3.8)
(93.4)
70.3
579
1628
780
965
10.7
2057
(8.2)
(26.2)
(73.8)
(35.3)
(43.7)
(3.8)
(93.2)
70.0
483
2021
631
773
10.8
1675
(8.0)
(27.1)
(72.9)
(35.3)
(43.2)
(3.8)
(93.6)
72.5
34
112
44
65
10.8
134
(9.5)§
(23.3)
(76.7)
(30.1)
(44.5)
(3.5)
(91.8)
500
1788
997
1308
518
141
1058
30.5
(18.0)
(64.5)
(36.0)
(47.2)
(18.7)
(5.2)
(38.5)
(6.9)
416
1421
795
1082
416
116
825
30.3
(18.9)
(64.5)
(36.0)
(49.0)
(18.9)
(5.3)
(37.7)
(6.9)
337
1150
654
891
338
96
668
30.6
(18.8)
(64.3)
(36.6)
(49.8)
(18.9)
(5.4)
(37.7)
(7.0)
25
96
48
67
27
9
55
28.7
(17.1)
(65.8)
(32.9)
(45.9)
(18.5)
(6.2)
(37.9)
(6.3)§
147.9 (16.3)
81.9 (10.7)
148.5 (16.3)
81.9 (10.7)
148.7 (16.3)
81.9 (10.7)
147.5 (16.9)
80.0 (10.0)§
147.1 (16.3)
81.5 (10.6)
147.7 (16.3)
81.6 (10.6)
147.9 (16.3)
81.5 (10.7)
146.4 (16.7)
79.6 (9.9)§
159.6
86.9
135.0
71.8
136.3
44.8
180.0
160.0
86.3
137.6
71.6
136.7
44.7
181.2
160.2
86.8
138.7
71.7
136.3
44.3
181.9
159.9
83.7
138.4
71.7
137.5
47.5
186.7
(11.1)
(10.9)
(65.4)
(20.9)
(38.9)
(14.5)
(140.4)
(11.1)
(11.0)
(67.3)
(21.2)
(39.6)
(14.4)
(137.2)
(11.1)
(10.8)
(67.7)
(20.9)
(39.5)
(14.3)
(137.0)
(13.9)
(10.2)
(70.7)
(21.1)
(38.0)
(15.6)§
(159.5)
ASCVD, atherosclerotic CVD; ECG, electrocardiogram; BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure; GFR, glomerular filtration rate; LDL,
low-density lipoprotein.
4Original prespecified HF outcome (treated, hospitalized, or fatal HF).
yValidated by any criteria set.
zNot validated by any criteria set.
§P b .05. Not validated compared to all HF.
tSee baseline characteristics in Ref. [10].
primary end point was a composite of nonfatal myocardial
infarction (MI) and fatal coronary heart disease (CHD). Major
secondary end points were all-cause mortality, stroke, combined
CHD (primary end point, coronary revascularization, hospitalized angina), and combined CVD (combined CHD, stroke, nonhospitalized treated angina, peripheral arterial disease, and HF).
ALLHAT enrolled men and women (47%), aged 55 years or
older (mean 67), with at least 1 additional CVD risk factor:
previous (N6 months) MI or stroke, left ventricular hypertrophy
(LVH) by electrocardiogram or echocardiogram, history of type
2 diabetes, current cigarette smoking, high-density lipoprotein
(HDL) b35 mg/dL, or other documented atherosclerotic CVD.
By race/ethnicity, 35% were black and 19% were Hispanic.
Individuals with a history of symptomatic HF and/or known
LVEF b35% were excluded. The protocol did not include a
study-measured LVEF.10
Validation Study Population
All ALLHAT participants with at least 1 reported hospitalization with a diagnosis of HF on admission or during their hospital
course were included in the Validation Study (Figure 1).
Diagnosis of HF in ALLHAT
HF diagnoses in ALLHAT were assigned by site physicians,
blinded to treatment assignment, guided by prespecified criteria
previously used in the SHEP, which required at least 1 symptom
and 1 sign of HF (Figure 2).6,16 Investigators were advised to be
conservative in diagnosing HF in patients with pulmonary
disease. Heart failure events identified by site physicians were
reported to the Clinical Trials Center (CTC) only if the patient
was also hospitalized or treated out of hospital, or if the event
was fatal. Because ALLHAT was a large simple trial, central
committee review of all HF events was not mandated by the
protocol. However, hospital discharges and death summaries
related to all end points, including HF, were required and were
reviewed by the CTC staff blinded to treatment assignments.
When HF events described in the documentation were not
reported as such by the site physician, or if the documentation
did not agree with the diagnosis, the CTC physician contacted
the site physician, who could then correct the report. To
further minimize underreporting, administrative databases
were searched regularly for potential events, which had to be
verified by site physicians to be included as end points.
American Heart Journal
January 2007
46 Einhorn et al
Data collection for the Validation Study
Between November 2001 and March 2002, the CTC
requested source documentation for all HF hospitalizations that
occurred between February 1, 1994, and March 31, 2002
(February 15, 2000, for the doxazosin/chlorthalidone comparison).17 It included face sheets with diagnoses and procedure
codes, emergency department notes, admission history and
physical examination, and cardiology and pulmonary consultation reports. Chest x-ray (CXR), echocardiography, cardiac
radionuclide study, magnetic resonance imaging, and cardiac
catheterization procedure reports were also requested, as well
as discharge summaries (if not previously received) and
autopsy reports. Each site submitted all obtainable records and
noted records that could not be obtained. Data collection for
the HF Validation Study ended in June 2002, approximately
1 month before any unblinding of study coordinators as to
treatment assignments in the lisinopril and amlodipine comparisons with chlorthalidone.
Diagnosis of HF in the Validation Study
Because of concerns about accessibility and quality of office
records, it was decided a priori that only hospitalized events
would be included as outcomes. Heart failure hospitalizations,
whether for initial presentation or after outpatient treatment,
constituted 78.9%, 78.6%, 74.5%, and 72.5% of all HF events in
the chlorthalidone-, amlodipine-, lisinopril-, and doxazosinbased groups, respectively. An abstracting form was developed
to provide documentation of the presence, absence, or
uncertainty/lack of information with respect to HF signs and
symptoms, past medical history, concurrent conditions, precipitating factors, cardiac function, and medications at discharge. Each individual event was independently abstracted by
2 blinded-to-treatment-assignment reviewers, randomly selected from a panel of 11 cardiology fellows from the Baylor
College of Medicine. To fully explore the validity of HF
diagnoses, the Validation Study used 6 different algorithms and
a global judgment of the reviewers.
Algorithms were based on ALLHAT and Framingham criteria.
The ALLHAT (A1) criteria (Figure 2), required a presence of
one sign and one symptom. Several more stringent ALLHATderived criteria sets were defined as follows: A2, excluded
events with z2+ pedal edema as the only sign; A3, required
CXR evidence of HF; A4, required intravenous treatment for HF
(diuretics, vasodilators, or positive inotropic agents), in
addition to meeting A1. The Framingham criteria (F1) require
2 major or 1 major and 2 minor criteria to make a diagnosis
(Figure 2).18 In the modified Framingham criteria (F2), the rule
is the same, but at least one criterion has to be diagnostic
(imaging, cardiac catheterization, or autopsy findings) and at
least one criterion clinical (Figure 2).
Algorithm-based diagnoses were assigned by computer.
In addition, the reviewers were asked whether, in their clinical
judgment, the patient had HF. The possible answers were byes,Q
bno,Q or bdon’t know.Q An event was validated if confirmed
by at least 1 reviewer, that is, considered to be HF by N2 of
the 3 observers, including the ALLHAT site physician.
Statistical methods
For each definition of HF, percent confirmation of site
physician–assigned diagnoses was calculated using contingen-
cy tables to determine the proportion of patients with at least
1 documented HF hospitalization for which the disease was
verified by at least 1 reviewer.
To determine the degree of support for the treatment effects
reported in the main result articles,10,11 analyses were repeated
for each Validation Study definition of HF to estimate RRs and
95% CIs. Relative risks were calculated from 2 2 tables
because the Cox proportional hazards principle was violated.
(The Cox proportional hazards regression model assumption
was examined using log-log plots and testing a treatment-bytime interaction term.)10 Heterogeneity of treatment effect
across racial subgroups (black vs nonblack) was examined by
stratification of the 2 2 tables using the Mantel-Haenszel
test of homogeneity. Because there were no significant
interactions ( P b .05), only overall RRs are reported. Life tables
were used to calculate mortality rates after the first validated
hospitalized event.
Analyses were conducted separately for the comparisons
of amlodipine or lisinopril versus chlorthalidone and doxazosin versus chlorthalidone because of different durations of
follow-up.10,11
Resultsg
For 2031 patients fulfilling the validation study
inclusion criteria, relevant hospital records were
requested for 2917 hospitalizations with a diagnosis of
HF on admission or during the hospital course. The CTC
received 2850 records for 1987 patients (98%), of which
97% (2778 records for 1935 patients) were suitable for
review (Figure 1).
Discharge summaries were received for 93% of first
events; admitting history and physical, and emergency
department notes for 71% and 38%, respectively. At least
1 type of imaging report was available for 79% of first
events (69%, 48%, 7%, 16%, respectively, for CXR,
echocardiography, radionuclide imaging, and cardiac
catheterization). For nonvalidated compared to validated
events, as would be expected, lesser documentation was
received for all validation criteria sets, especially emergency department notes, admitting history and physical,
and imaging reports (43%, 79%, 76% [CXR], and 61%
[other imaging] vs 30%, 55%, 56%, and 45%, respectively, for A1). Nursing notes and medication charts
were not requested. Information on intravenous medications was missing for 27% of reviewed events,
whereas absence of intravenous treatment was documented only in 3%.
Verification of ALLHAT site physician-assigned
diagnoses
Percent agreements with site physician–assigned
diagnoses for the various Validation Study criteria for HF
g
Unless noted otherwise, results are presented for the
chlorthalidone/amlodipine/lisinopril comparison and are similar
to the chlorthalidone/doxazosin comparison.
American Heart Journal
Volume 153, Number 1
Einhorn et al 47
Table II. Incident hospitalized HF outcomes by antihypertensive treatment group (amlodipine/lisinopril vs chlorthalidone)
Chlorthalidone (n = 15 255)
No. of
events
Site physician
ALLHAT (A1)
ALLHAT (A2)
ALLHAT (A3)
ALLHAT (A4)
Framingham (F1)
Framingham (F2)
Reviewers
617
426
423
406
344
489
442
507
6-y Rate per
100 persons
(SE)
5.7
4.1
4.1
3.9
3.3
4.7
4.4
4.9
(0.26)
(0.23)
(0.23)
(0.23)
(0.21)
(0.24)
(0.24)
(0.25)
Amlodipine (n = 9048)
Incidence rate/
1000 patient-years
(95% CI)
8.7
6.0
5.9
5.7
4.8
6.9
6.2
7.1
(8.0-9.4)
(5.4-6.6)
(5.4-6.5)
(5.2-6.3)
(4.3-5.4)
(6.3-7.5)
(5.7-6.8)
(6.5-7.8)
were similar across randomized treatment groups
(Figure 1). The highest (84%) was for reviewers’
judgment (Reviewers), followed by F1 (80%). The
percent agreement for A2 was virtually the same as that
for A1 because there were only 12 patients with z2+
ankle edema as the only prespecified sign of HF. All
events validated by Reviewers met at least one other set
of criteria. In 8% of patients, both reviewers concluded
that they were uncertain about the diagnosis, and in
another 8%, both reviewers agreed that documentation
did not support the diagnosis of HF. Ninety-two percent
of site physician–assigned diagnoses met at least 1 set of
the validation criteria, and two thirds met both A1 and
F1 criteria.
Patients whose diagnoses were confirmed by the
Validation Study were similar to all patients who
developed HF, whether hospitalized or treated without
hospitalization (Table I). The 146 patients in whom HF
was not confirmed were slightly older, had lower body
mass index and diastolic blood pressure, and had higher
HDL cholesterol at baseline. Of events validated by the
algorithms, Reviewers agreed with 81.3%, 81.4%, 83.0%,
86.0%, 76.7%, and 81.3% of diagnoses for A1 to A4, F1,
and F2, respectively.
Verification of treatment effects
As shown in Table II and Figure 3, RRs for
developing HF, when verified by review using any
criteria in the Validation Study, were all higher than
those reported for hospitalized and fatal events in the
main result articles, and for the amlodipine and
doxazosin comparisons with chlorthalidone, they
remained highly statistically significant ( P b .0001)
despite lower event rates.10,11 For the lisinopril/chlorthalidone comparison, the RRs ranged from 1.21 for
the A3 criterion (requiring CXR evidence of HF) to
1.12 for F2. The lower limits of the 95% CIs were all
close to unity, but for 4 criteria sets (A1, A2, A3, and
Reviewers), the RRs were statistically significant.
No. of
events
508
369
365
348
295
412
369
436
6-y Rate per
100 persons
(SE)
7.9
5.8
5.7
5.4
4.6
6.3
5.8
6.7
(0.40)
(0.35)
(0.34)
(0.34)
(0.31)
(0.35)
(0.34)
(0.36)
Incidence rate/
1000 patient-years
(95% CI)
12.1
8.7
8.6
8.2
7.0
9.8
8.7
10.4
(11.1-13.2)
(7.9-9.7)
(7.8-9.6)
(7.4-9.1)
(6.2-7.8)
(8.9-10.8)
(7.9-9.7)
(9.4-11.4)
Incidence rates and mortality after validated
hospitalized HF events
Validated hospitalized HF incidence rates per
1000 patient-years ranged from 8.3 for Reviewers to 5.6
for A4 in the amlodipine/lisinopril/chlorthalidone comparison and from 9.3 for Reviewers to 6.1 for A4 in the
doxazosin/chlorthalidone comparison.
All-cause mortality rates after incident HF hospitalization were similar for validated compared to nonvalidated
events (Figure 4). Percentage of fatal among first HF
events validated by any of the criteria sets (8.7%) was
considerably lower compared to first events not validated by any of the criteria sets (31.5%). (Validation rates
for fatal events were only 47%, 62%, 48%, and 65% for
A1, F1, F2, and Reviewers, respectively.)
Discussion
Principal findings and comparisons to prior reports
The primary objective of this central review of HF
events in ALLHAT was to evaluate the HF diagnoses
assigned by ALLHAT site physicians by using independent reviewers and prespecified criteria to determine
the degree of support for the trial’s conclusions with
regard to the HF outcome. Our major findings were as
follows: (1) the site physicians’ diagnoses of HF were
confirmed in most patients, according to 6 prespecified
algorithms and reviewers’ judgment; (2) the previously
reported treatment differences based on ALLHAT
investigators’ reports were corroborated by RRs calculated when applying the 7 criteria sets; (3) 6-year
incidence rates of validated hospitalized HF were
comparable in magnitude to those of stroke (5.6%)
and to about half of those of combined nonfatal MI
and CHD deaths (11.4%),10 demonstrating the clinical
significance of this end point in the hypertensive
population. The magnitude of the mortality rates
subsequent to hospitalized HF (55% at 5 years) further
underscores the importance of prevention.
American Heart Journal
January 2007
48 Einhorn et al
Table II. continued
Lisinopril (n = 9054)
No. of
events
415
298
295
291
231
328
295
347
6-y Rate per
100 persons
(SE)
6.3
4.6
4.6
4.5
3.5
5.0
4.6
5.2
(0.35)
(0.30)
(0.30)
(0.29)
(0.26)
(0.31)
(0.30)
(0.31)
Incidence rate/
1000 patient-years
(95% CI)
10.0
7.1
7.0
6.9
5.5
7.8
7.0
8.3
(9.1-11.0)
(6.4-8.0)
(6.3-7.9)
(6.2-7.8)
(4.8-6.3)
(7.0-8.7)
(6.3-7.9)
(7.5-9.2)
Amlodipine vs chlorthalidone
Lisinopril vs chlorthalidone
RR (95% CI)
from 2 2 table
RR (95% CI)
from 2 2 table
1.39
1.46
1.46
1.45
1.45
1.42
1.41
1.45
The high confirmation rates of the site physician
diagnoses by independent reviewers and by Framingham
criteria are similar to those reported by other studies.19,20
Roger et al4 reported that in Olmstead County, Minnesota,
82% of the cases coded by clinicians as ICD 428 met
Framingham criteria for HF, a number similar to the 80%
agreement found in this study. They also compared cases
coded by Olmstead County clinicians as ICD 428 with
their reviewers’ clinical judgment and found 90% confirmation rate. Again, this rate is similar to the 84%
agreement with the reviewers’ judgment in ALLHAT.
The comparisons of treatment effects for amlodipineand doxazosin-based strategies with a chlorthalidonebased strategy for the 7 criteria sets examined provide
strong and conclusive evidence for the superiority of a
chlorthalidone-based therapy in preventing transition
from hypertension to overt HF. An excess of HF events
was observed even when A2 criteria that exclude events
with z2+ pedal edema as the only prespecified sign of
HF were used. For the lisinopril/chlorthalidone comparison, the excess risk of HF with lisinopril was less
dramatic but larger than that found for hospitalized
events reported by ALLHAT investigators and statistically
significant using reviewers’ judgment and ALLHAT
criteria (A1-A3). As the most conservative interpretation,
the results of the Validation Study provide strong and
conclusive evidence that an ACE inhibitor did not offer
superior to chlorthalidone benefit in preventing HF.
They also strengthen the published main trial result with
respect to the post hoc end point of hospitalized and
fatal HF (RR 1.10, 95% CI 0.98-1.23) and support our
earlier conclusion for the ALLHAT prespecified HF end
point (treated out of hospital, hospitalized, or fatal) that
chlorthalidone-based treatment is superior to lisinoprilbased in preventing new onset HF in both blacks and
nonblacks (RR 1.19, 95% CI 1.07-1.31).10,21
ALLHAT findings with respect to the ACE inhibitor
seemed surprising given results of placebo-controlled
trials.7,22-24 However, SHEP investigators reported a 49%
(1.24-1.56)
(1.27-1.68)
(1.27-1.67)
(1.26-1.67)
(1.24-1.69)
(1.25-1.62)
(1.23-1.61)
(1.28-1.64)
P
b.0001
b.0001
b.0001
b.0001
b.0001
b.0001
b.0001
b.0001
1.13
1.18
1.18
1.21
1.13
1.13
1.12
1.15
(1.00-1.28)
(1.02-1.36)
(1.02-1.36)
(1.04-1.40)
(0.96-1.33)
(0.99-1.30)
(0.97-1.30)
(1.01-1.32)
P
.04
.03
.03
.01
.14
.08
.11
.04
( P b .001) decrease in the risk of HF with chlorthalidone
versus placebo.6 For ACE inhibitors, this reduction was
only 20% ( P b .001) in SOLVD Prevention (captopril),
23% ( P b .001) in HOPE (ramipril), and 25% ( P = .02) in
PEACE (trandolapril).22-24
The results of drug comparison trials are mixed but,
overall, not inconsistent with the ALLHAT findings.7,25
Notably, only one trial directly compared a diuretic with
an ACE inhibitor.26 The ANBP2 study, which compared a
diuretic-based treatment (hydrochlorothiazide recommended, dose not stated) with an ACE inhibitor (enalapril
recommended, dose not stated), was an open-label
practice-based trial in 6083 hypertensive participants
(compared with 24 301 in the 2 similar arms in ALLHAT).
The HF rates (not stated whether based on all or
hospitalized events only) were not statistically different
between the 2 first-step drugs (RR 0.85, 95% CI 0.62-1.18),
and the upper limit of the 95% CI suggests that the HF
results of the 2 trials may be consistent.27 It is also possible
that ANBP2 site clinicians, who were aware of treatment
assignments, were more conservative in diagnosing HF
in those treated with an ACE inhibitor. Similarly, in the
open-label ASCOT comparing an amlodipine-based
regimen with an atenolol-based regimen, physicians
could have diagnosed HF less frequently in the amlodipine-based treatment arm by not considering peripheral
edema as a potential symptom of HF, especially in an
outpatient setting.25 This diagnostic/reporting bias is not
an issue in ALLHAT where site clinicians were blinded to
treatment assignments.
In ALLHAT, observed HF differences were larger
earlier in the follow-up. Relative risks (95% CIs) for year
1 of follow-up, compared with subsequent years, were
respectively 2.08 (1.58-2.74) versus 0.96 (0.85-1.10) and
2.22 (1.69-2.91) versus 1.22 (1.08-1.38) for the lisinopril
and amlodipine comparisons with chlorthalidone, respectively. There was no significant interaction between
prior medication use and treatment group, and data on
the type of prior antihypertensive medication collected
American Heart Journal
Volume 153, Number 1
Einhorn et al 49
Figure 3
A, Cumulative hospitalized HF incidence rates reported by site physicians and validated by ALLHAT (A1), Framingham (F1), and Reviewers
criteria sets (amlodipine/lisinopril/chlorthalidone). B, Cumulative hospitalized HF incidence rates reported by site physicians and validated by
ALLHAT (A1), Framingham (F1), and Reviewers criteria sets (doxazosin/chlorthalidone).
at the trial close-out provide no evidence that discontinuation of prior medication, for example, diuretics, at
the time of randomization contributed to the 2-fold
increase in the incidence of HF hospitalizations in the
ACE inhibitor or calcium-channel blocker group during
year 1.28
Various explanations have been entertained to explain
the ALLHAT HF results, including over- and misdiagnosis
of HF by ALLHAT physician-investigators, a diverse
group of clinicians in a variety of practice settings. We
believe that this manuscript provides conclusive evidence that ALLHAT findings of chlorthalidone-based
treatment’s superiority over amlodipine-, lisinopril- and
doxazosin-based strategies in preventing the transition
from hypertension to symptomatic HF were not due to
the misdiagnosis of HF.
The ALLHAT protocol did not include a measurement
of LVEF either before or at the time of HF diagnosis.
Information on LVEF was collected as part of this
Validation Study, and a report on these data is in
preparation. If the superiority of diuretics is limited to
preventing HF in a setting of preserved LVEF, this could
explain the equivalent results beyond year 1 of follow-up
with interim MIs contributing to a larger number of
individuals with impaired LVEF. Phenotypic and prognostic correlates of the various diagnostic algorithms
American Heart Journal
January 2007
50 Einhorn et al
Figure 3 (continued)
will also be explored. Small differences in RRs among
the criteria sets may be related to preferential identification of HF with impaired LVEF by, for example, F2
(EFb35% is a major criterion) or A4 (preferential
documentation on the use of intravenous pressors
compared with diuretics).
Strengths and limitations
It is important to emphasize that HF was not a primary
end point in ALLHAT, and therefore, only limited
information was collected prospectively for these
events. The Validation Study was conducted retrospectively during the trial close-out, but information was
available for most patients, and record collection was
completed before unblinding any of the study coordi-
nators. In addition, the evaluation was rigorous and
conducted by reviewers unrelated to the study and
blinded to treatment assignments. To minimize underreporting of the events, all hospital discharge summaries
were reviewed by the CTC staff blinded to treatment
assignment, and administrative databases were searched
for potential events. Furthermore, reporting bias is not
of concern because ALLHAT site physicians who were
assigning the diagnoses were blinded to treatment
assignments. Although for feasibility reasons and given
known limitations of outpatient records this central
review was limited to hospitalized events only, it
included most hospitalized HF events in ALLHAT.
Consequently, it is highly unlikely that a prospective
adjudication by the end points committee would have
American Heart Journal
Volume 153, Number 1
Einhorn et al 51
Figure 4
A, All-cause mortality after hospitalization with first HF event; incident fatal events not included (amlodipine/lisinopril/chlorthalidone). B, Allcause mortality after hospitalization with first HF event; incident fatal events not included (doxazosin/chlorthalidone).
been more thorough, of higher quality, or more
systematic than this retrospective review. Additional
clinical data abstracted during the review of the 2778 HF
events will provide information for future publications
on clinical presentation and circumstances surrounding
transition from hypertension to overt HF in these
1935 ALLHAT participants.
Implications for future clinical trials
Heart failure proved to be a common outcome in
ALLHAT and one that was affected differentially by the
randomized treatment assignments. In addition, patients
who developed HF had significantly poorer survival than
those who did not. Whether this poor prognosis can be
altered presents an important clinical and public health
question. Thus, in planning future hypertension treatment trials (and perhaps also treatment trials in other
populations at high risk of HF, such as diabetics and
survivors of acute coronary syndromes),29,30 serious
consideration should be given to including HF in the
primary end point, along with death, MI, and stroke.
In ALLHAT, near equivalent results for treatment
effects on the HF outcome were obtained whether using
a more specific end point of validated hospitalized HF
(RRs between 1.12 and 1.21 across criteria sets) or a
broader end point based on site physicians’ reports of
American Heart Journal
January 2007
52 Einhorn et al
Figure 4 (continued)
unadjudicated hospitalized, fatal, or treated out-of-hospital HF (RR 1.19, 95% CI 1.07-1.31). Financial considerations make the latter an attractive possibility in
trial design.
The experience from the ALLHAT Validation Study
and from other community-based studies indicates that
community physicians can reliably identify HF
events.4,19 In ALLHAT, about one fourth of the patients
with HF were never hospitalized, including those
treated in an outpatient setting during the last year of
follow-up. Whether identification of the incident HF
event occurs in an outpatient setting or in a hospital
depends on many circumstances, including clinical
presentation, access to care, and patient health care–
seeking behaviors. ALLHAT experience showed that in
a diverse demographic and medical setting, limiting the
HF outcome to hospitalized events may result in less
precise relative risk estimates and insufficient power to
detect a treatment difference.
Implications for clinical practice
The prevention of HF is an important goal of
antihypertensive management in clinical practice.
New-onset HF is a relatively common complication of
hypertension with high subsequent mortality. The
results of the ALLHAT Validation Study confirm that
thiazide-type diuretics should be the preferred first-step
drug therapy for prevention of HF in high-risk patients
with hypertension. This indication is further strengthened by the two-fold higher risk of HF for lisinopril and
amlodipine compared with chlorthalidone during year 1
of treatment.28
American Heart Journal
Volume 153, Number 1
We thank Dr Jeffrey Cutler, ALLHAT Project Office
Director, for his leadership in the design, conduct, and
reporting of ALLHAT and for shepherding the Heart
Failure Validation Study from its conceptual inception
through data collection and reporting. We also thank
the members of the ALLHAT Steering Committee10 for
their approval and support of this endeavor and for
their leadership in assuring its successful implementation; the reviewers; and Dr Gabriel Habib, Director
of Education at the Michael E. DeBakey VA Medical
Center, Houston, TX.
References
1. Massie BM, Shah NB. Evolving trends in the epidemiologic factors of
heart failure: rationale for preventive strategies and comprehensive
disease management. Am Heart J 1997;133:703 - 12.
2. Hunt SA, Baker DW, Chin MH, et al. ACC/AHA Guidelines for the
evaluation and management of chronic heart failure in the adult:
executive summary. Circulation 2001;104:2996 - 3007.
3. Levy D, Kenchaiah S, Larson MG, et al. Long-term trends in the
incidence of and survival with heart failure. N Engl J Med
2002;347:1397 - 402.
4. Roger VL, Weston SA, Redfield MM, et al. Trends in heart failure
incidence and survival in a community-based population. JAMA
2004;292:344 - 50.
5. Levy D, Larson MG, Vasan RS, et al. The progression from
hypertension to congestive heart failure. JAMA 1996;275:1557 - 62.
6. Kostis JB, Davis BR, Cutler J, et al. Prevention of heart failure by
antihypertensive drug treatment in older persons with isolated
systolic hypertension. SHEP Cooperative Research Group. JAMA
1997;278:212 - 6.
7. Blood Pressure Lowering Treatment Trialists’ Collaboration. Effects of
different blood-pressure-lowering regimens on major cardiovascular events: results of prospectively-designed overviews of randomized trials. Lancet 2003;362:1527 - 35.
8. Psaty BM, Lumley T, Furberg CD, et al. Health outcomes associated
with various antihypertensive therapies used as first-line agents.
JAMA 2003;289:2534 - 44.
9. Davis BR, Cutler JA, Gordon DJ, et al. for the ALLHAT Research
Group. Rationale and design for the Antihypertensive and Lipid
Lowering Treatment to Prevent Heart Attack Trial (ALLHAT). Am J
Hypertens 1996;9:342 - 60.
10. The ALLHAT Officers and Coordinators for the ALLHAT Collaborative Research Group. Major outcomes in high-risk hypertensive
patients randomized to angiotensin-converting enzyme inhibitor or
calcium channel blocker vs diuretic: the Antihypertensive and LipidLowering Treatment to Prevent Heart Attack Trial (ALLHAT). JAMA
2002;288:2981 - 97.
11. The ALLHAT Officers and Coordinators for the ALLHAT Collaborative Research Group. Diuretic versus alpha-blocker as first-step
antihypertensive therapy: final results from the Antihypertensive and
Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT).
Hypertension 2003;42:239 - 46.
12. Weber MA. The ALLHAT report: a case of information and
misinformation. J Clin Hypertens 2003;5:9 - 13.
13. Messerli FH. ALLHAT, or the soft science of the secondary end point.
Ann Intern Med 2003;139:777 - 80.
Einhorn et al 53
14. Davis BR, Furberg CD, Wright JT, et al. for the ALLHAT
Collaborative Research Group. ALLHAT: setting the record straight.
Ann Intern Med 2004;141:39 - 6.
15. Staessen JA, Birkenhager WH. Evidence that new antihypertensives
are superior to older drugs. Lancet 2005;366:869 - 71.
16. Piller LB, Davis BR, Cutler JA, et al. Validation of heart failure events
in ALLHAT participants assigned to doxazosin. Am J Hypertens
2001;14:180A.
17. The ALLHAT Officers and Coordinators for the ALLHAT Collaborative Research Group. Major cardiovascular events in hypertensive
patients randomized to doxazosin vs chlorthalidone: the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack
Trial (ALLHAT). JAMA 2000;283:1967 - 75.
18. Ho KKL, Anderson KM, Kannel WB, et al. Survival after the onset of
congestive heart failure in Framingham Heart Study subjects.
Circulation 1993;88:107 - 15.
19. Peyster E, Norman J, Domanski M. Prevalence and predictors of
heart failure with preserved systolic function: community hospital
admissions of a racially and gender diverse elderly population. J
Card Fail 2004;10:49 - 4.
20. Schellenbaum GD, Rea TD, Heckbert SR, et al. Survival associated
with two sets of diagnostic criteria for congestive heart failure. Am J
Epidemiol 2004;160:628 - 35.
21. Wright JT, Dunn JK, Cutler JA, et al. Outcomes in hypertensive black
and non-black patients treated with chlorthalidone, amlodipine, and
lisinopril. JAMA 2005;293:1595 - 608.
22. The SOLVD Investigators. Effect of enalapril on mortality and
the development of heart failure in asymptomatic patients with
reduced left ventricular ejection fractions. N Engl J Med
1992;327:685 - 91.
23. Yusuf S, Sleight P, Pogue J, et al. Effects of an angiotensinconverting-enzyme inhibitor, ramipril, on cardiovascular events in
high-risk patients. The Heart Outcomes Prevention Evaluation Study
Investigators. N Engl J Med 2000;342:145 - 53.
24. Braunwald E, Domanski MJ, Fowler SE, et al. Angiotensin-converting-enzyme inhibition in stable coronary artery disease. N Engl J
Med 2004;351:2058 - 68.
25. Dahlof B, Sever PS, Poulter NR, et al. for the ASCOT investigators.
Prevention of cardiovascular events with an antihypertensive
regimen of amlodipine adding perindopril as required versus
atenolol adding bendroflumethiazide as required, in the AngloScandinavian Cardiac Outcomes Trial-Blood Pressure Lowering
Arm (ASCOT-BPLA): a multicentre randomised controlled trial.
Lancet 2005;366:895 - 06.
26. Wing LMH, Reid CM, Ryan P, et al. for the Second Australian
National Blood Pressure Study Group. A comparison of
outcomes with angiotensin-converting-enzyme inhibitors and diuretics for hypertension in the elderly. N Engl J Med 2003;348:
583 - 92.
27. Cutler JA. The ANBP2 and ALLHAT: conflicting or consistent? J Clin
Hypertens 2003;5:192 - 5.
28. Davis BR, Piller LB, Cutler JA, et al. Role of diuretics in the prevention of
heart failure: the antihypertensive and lipid-lowering treatment to
prevent heart attack trial. Circulation 2006;113:2201 - 10.
29. Lewis EF, Moye LA, Rouleau JL, et al. Predictors of late development
of heart failure in stable survivors of myocardial infarction: the CARE
study. J Am Coll Cardiol 2003;42:1446 - 53.
30. Yki-Jarvinen H. The PROactive study: some answers, many questions. Lancet 2005;366:1241 - 2.